Modelpublic

oumoumad/ltx-2.3-dearchive-lora

sourceHugging Faceotherupdated 5mo agoView on Hugging Face

45likes

Model Card

dearchive · LTX-2.3 IC-LoRA

An In-Context LoRA for LTX-2.3 (dev, 22B) that takes real archive video — old broadcast B&W footage, low-res low-bitrate web rips, sepia-toned silent-era prints — and rewrites it as if shot more recently (colored, high definition, sharp detail, modern cinematography). Tested on actual archive footage, not just synthetic equivalents.


Base model	`Lightricks/LTX-2.3` (`ltx-2.3-22b-dev.safetensors`)
Strategy	`video_to_video` (in-context, reference-conditioned)
LoRA rank / alpha	128 / 128
Trainable params	855,638,016 (~0.86 B)
Optimizer	Prodigy (D-Adaptation), `lr=1.0`, bias-correction + safeguard-warmup
Scheduler	cosine
Mixed precision	bf16 + int8-quanto
Reference downscale	1 (full res)
Resolution buckets	`960×544×97; 960×544×49`
Steps	5000
Save interval	every 500 steps
Seed	42

What it learns to undo

Real archive YouTube uploads of mid-20th-century broadcast footage (Bruce Lee interviews, Chaplin web rips, etc.) are dominated by resolution + compression loss, not silent-era film damage. The training pipeline mirrors that:

clean 1920×1080
   → tonal degrade  (B&W via Rec.601 luma, optional family tint, contrast/gamma)
   → capture-σ blur (tier-scaled, simulates lens / multi-gen optical printing)
   → downscale to 360p / 270p / 240p (bilinear)
   → low-bitrate h264 encode @ 60–320 kbps
   → optional re-encode 1–3 generations (compounds compression artifacts)
   → optional hqdn3d denoise (heavy tier only)
   → Lanczos upscale back to 1920×1080  (matches inference-time user upscale)

Three corruption families are sampled per pair:

Family	What it matches	Calibration ref
`chain_neutral`	neutral B&W broadcast tier	Bruce Lee Philosophy (yt nzQWYHHqvIw, 640×360 / 62 kbps)
`tint_tape`	cool-green VHS-tape oxidation	Bruce Lee Nunchucks (yt qHe6vhexm6g, 320×240 / 88 kbps)
`tint_sepia`	warm-brown film-age fade	Safety Last (1923, sepia mid-tones)

Tape family gets the heavy chain at the heaviest tier (smashed BL-Nunchucks-class output); neutral and sepia families use 0.65 / 0.70 multipliers on the capture σ and 0.5× the denoise probability so they preserve the gentler mid-tier character.

Dataset

—53 source clips, landscape ≥720p, ≥6 s
—3 corrupted variants per source → 159 pairs total (151 train + 8 held-out validation)
—All target/reference at 1920×1080 16:9 (matching aspects — this LoRA does not outpaint)
—Frames: 97 frames @ 24 fps (4.04 s; LTX-2 requires n % 8 == 1)
—Caption (single, generic): "A modern, high-resolution video shot in vivid color (or natural monochrome), with sharp detail, clean tonality, and contemporary cinematography."

Files

File	Step
`lora_weights_step_05000.safetensors`	5000 (final)
`training_state_step_05000.pt`	5000
`DeArchive.json`	example ComfyUI workflow

Quick inference

bash

git clone https://github.com/Lightricks/LTX-2.git && cd LTX-2 && uv sync
uv pip install peft

uv run python packages/ltx-trainer/scripts/inference.py \
    --checkpoint /path/to/ltx-2.3-22b-dev.safetensors \
    --text-encoder-path /path/to/gemma-3-12b-it-qat-q4_0-unquantized \
    --lora-path /path/to/lora_weights_step_05000.safetensors \
    --reference-video /path/to/your_lanczos_upscaled_archive.mp4 \
    --prompt "A modern, high-resolution video shot in vivid color, sharp detail, contemporary cinematography." \
    --width 960 --height 544 --num-frames 97 --frame-rate 24 \
    --num-inference-steps 50 --guidance-scale 4.0 \
    --output dearchive_restored.mp4